The technique of condition monitoring and health management is the breakthrough key technology of the domestic civil aircraft operators support. With civil aircraft bleed air system as the object of the research, the article designs an anomaly detection process to monitor the health of civil aircraft bleed air system and proposes the fault detection algorithm of bleed air system. According to the operating principle and operating mode of bleed air system, extracts 9th stage and 5th stage regulator corresponding monitoring data from quick access recorder(QAR) decoded data. Exponentially weighted moving average(EWMA) control chart method by introduction of accuracy is optimized, false alarm rate and missing alarm rate. The smoothing parameter, number of samples to monitor the data points and the coefficient of control limit for EWMA control chart can be determined based on the optimization index. The method is validated by using actual data collected in airline. The results show that, comparing to the current airline programs, the proposed method not only can reduce the false alarm rate of failure warning, improve accuracy, but also found fault ahead of time, which may provide support for the airline to arrange aircraft maintenance programs reasonably, reduce aircraft downtime.